NASA Technical Reports Server (NTRS) 20110011920: Canadian Experiment for Soil Moisture in 2010 (CanEX-SM10): Overview and Preliminary Results

1.6 MB·English

by  NASA Technical Reports Server (NTRS)

IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING 1
Canadian Experiment for Soil Moisture in
2010 (CanEX-SM10): Overview and
Preliminary Results
Ramata Magagi, Aaron Berg, Kalifa Goïta, Stephane Belair, Tom Jackson, Brenda Toth, Anne
Walker, Heather McNairn, Peggy O’Neill, Mahta Moghddam, Imen Gherboudj, Andreas Colliander,
Michael Cosh, John Belanger, Mariko Burgin, Josh Fisher, Sab Kim, Louis-Philippe B.-Rousseau,
Nadjib Djamaï, Jiali Shang, and Amine Merzouki
and the airborne L-band radiometer data exhibited spatial and
Abstract— The Canadian Experiment for Soil Moisture in 2010 temporal variability and polarization dependency. The temporal
(CanEx-SM10) was carried out in Saskatchewan, Canada from evolution of SMOS soil moisture product matched that observed
31 May to 16 June, 2010. Its main objective was to contribute to with the ground data, but the absolute soil moisture estimates did
Soil Moisture and Ocean salinity (SMOS) mission validation and not meet the accuracy requirements (0.04 m3/m3) of the SMOS
the pre-launch assessment of Soil Moisture and Active and mission. AMSR-E soil moisture estimates are more closely
Passive (SMAP) mission. During CanEx-SM10, SMOS data as correlated with measured soil moisture.
well as other passive and active microwave measurements were
collected by both airborne and satellite platforms. Ground-based Index Terms— SMOS, soil moisture, brightness temperature,
measurements of soil (moisture, temperature, roughness, bulk validation, field experiment, agricultural and boreal forested
density) and vegetation characteristics (Leaf Area Index, areas, microwave airborne sensors, AMSR-E.
biomass, vegetation height) were conducted close in time to the
airborne and satellite acquisitions. Besides, two ground-based in
situ networks provided continuous measurements of I. INTRODUCTION
meteorological conditions and soil moisture and soil temperature

profiles. Two sites, each covering 33 km x 71 km (about two
SMOS pixels) were selected in agricultural and boreal forested Remote sensing of soil moisture is a key component of
areas in order to provide contrasting soil and vegetation several observing and research programs including the
conditions. This paper describes the measurement strategy, Global Energy and Water Cycle Experiment (GEWEX),
provides an overview of the data sets and presents preliminary the Integrated Global Water Cycle Observation (IGWCO) and
results. Over the agricultural area, the airborne L-band the Global Soil Wetness Project (GSWP), among others. This
brightness temperatures matched up well with the SMOS data.
is related to the fact that soil moisture plays a critical role in
The Radio frequency interference (RFI) observed in both SMOS
governing global water and energy cycles. Recently, Jung et
al. [1] linked the decline in the global evapotranspiration since
Manuscript received April 8, 2011. This work was supported by several 1998 to a potential limitation in soil moisture supply. At
agencies in Canada (Natural Sciences and Engineering Research Council of
regional and local scales, soil water availability affects the
Canada, Environment Canada, Canadian Space Agency, and Agriculture and
distribution of vegetation and crop health, and impacts flood
Agri-Food Canada) and USA (National Aeronautics and Space
Administration, United States Department of Agriculture). risk. Bélair et al. [2] and Koster et al. [3] have shown the
R. Magagi, K. Goïta, I. Gherboudj, L.-P. Rousseau and N. Djamaï are with importance of the initial soil moisture state for improved
the Université de Sherbrooke, Département de géomatique appliquée,
climate and weather forecasts while Berg and Mulroy [4] have
Sherbrooke, QC J1K 2R1, Canada ([email protected]).
demonstrated the value of modeled soil moisture for
A.A. Berg and John Belanger are from the University of Guelph,
Department of Geography, Guelph, ON N1G 2W1, Canada improving streamflow forecasts. Numerous microwave
([email protected]). satellite missions (RADARSAT-2, AMSR-E, ALOS-PalSAR,
B. Toth is from Environment Canada MSC Hydrometeorology and Arctic etc.) currently provide data which can be used to estimate and
Lab, Saskatoon, (SK) Canada
monitor changes in soil moisture. In addition, the European
S. Belair is from Environment Canada, Meteorological Research Branch,
2121, Trans-Canada Highway, Dorval, QC H9P 1J3, Canada Space Agency’s (ESA) new mission Soil Moisture and Ocean
A. Walker is from Environment Canada, Climate Research Division, 4905 Salinity (SMOS) and the National Aeronautics and Space
Dufferin Street, Toronto, ON M3H 5T4, Canada Administration’s (NASA) future mission Soil Moisture Active
H. McNairn, A. Merzouki, and J. Shang are from Agriculture and Agri-
and Passive (SMAP) are dedicated to monitoring global soil
Food Canada, 960 Carling Ave, Ottawa, ON K1A 0C6.
M. Moghddam, and M. Burgin are from the University of Michigan, moisture information [5], [6]. Exploitation of this new satellite
Electrical Engineering and Computer Science Department, Ann Arbor, MI microwave data requires intensive campaigns to collect
48109-2122, USA. ground and airborne data to validate SMOS brightness
S. Kim, A. Colliander, P. O’Neill, and J. Fisher are from Jet Propulsion
temperature data and soil moisture products. Lessons learned
Laboratory, Pasadena, CA 91109, USA.
T. Jackson, and M. Cosh are from USDA-ARS Hydrology and Remote from SMOS investigations, especially when complemented
Sensing Lab, Beltsville, MD 20705, USA.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING 2
with airborne radar data, will contribute to SMAP algorithm II. SITES
development and validation.
2.1. General description of the study sites
Several international field experiments, over a variety of
landscapes, have been devoted to the assessment of SMOS The CanEx-SM10 experiment was conducted over two
brightness temperature data and soil moisture products. Each disparate landscapes including an agricultural and a forested
of these has value to the overall assessment of the SMOS region of Saskatchewan, Canada (Fig. 1). Both the agricultural
products. The Canadian Experiment for Soil Moisture in 2010 Kenaston site and the forested site of the Boreal Ecosystem
(CanEx-SM10) complements these by focusing on a different Research and Monitoring Sites (BERMS) covered an area of
climate region. Details on these field campaigns can be found 33 km x 71 km (about two SMOS pixels). These sites were
in [7]-[12]. selected to minimize as much as possible large lakes, which
The purpose of this paper is to present an overview of can be problematic for the validation of coarse resolution
CanEx-SM10 [11] which took place from 31 May to 16 June microwave data. In addition, the following aspects were
2010 in Saskatchewan, Canada. CanEx-SM10 was a considered during the selection of the two study sites:
collaborative effort between researchers in Canada and the • Both the Kenaston and BERMS sites benefit from long term
U.S.A. The campaign was designed to collect field in situ soil moisture measurement networks managed by EC
measurements for the validation of SMOS data, the pre-launch at BERMS and by EC and U of G at Kenaston.
assessment of SMAP soil moisture products, and the Meteorological stations are also available.
development of soil moisture retrieval algorithms specifically • The cropping system present within the Kenaston site is
for agricultural and boreal forest areas in Saskatchewan, very typical of the Canadian Prairie region, consisting of
Canada. To meet the abovementioned objectives, L-band cereal, oilseed and pasture-forage crops. Fields in this
passive microwave data were acquired with a radiometer region are considered large, reaching 60 ha in size. The
mounted on a Twin Otter aircraft owned by the National cropping mix and field sizes of the Kenaston area are well
Research Council of Canada (NRC) and managed by suited for testing the retrieval algorithms of soil and
Environment Canada. Data were also acquired by NASA’s vegetation parameters from microwave remote sensing.
Uninhabited Aerial Vehicle Synthetic Aperture Radar • The BERMS site benefits from long term ecological data
(UAVSAR), a polarimetric L-Band synthetic aperture radar collected during previous research programs such as the
(SAR) sensor flown on a Gulfstream-III aircraft. Coincident Boreal Ecosystem-Atmosphere Study (BOREAS in 1994
with airborne and satellite (SMOS, AMSR-E, RADARSAT-2 and 1996) and BERMS (1996 to present).
and ALOS-PalSAR) acquisitions, ground measurements were
made to characterize the soil (moisture, temperature,
roughness, bulk density) and the vegetation (height, biomass,
Leaf Area Index (LAI), density). In addition, two ground-
based networks managed by the University of Guelph (U of
G) and Environment Canada (EC) provided continuous
measurements of soil moisture over the study area. At the time
of the present study, SMOS is in its early operational phase
(since June 2010) and consequently, the large dataset
collected during CanEx-SM10 provides critical data to
improve the soil moisture retrieval algorithms for both
agricultural and boreal forest landscapes, and to develop
downscaling approaches. The large agricultural fields
(approximately 60 ha), characteristic of Canada’s Prairies
region, are well suited to address L-band coarse resolution
passive microwave research questions. CanEx-SM10 was the
first attempt in Canada to acquire soil moisture observations
simultaneous with satellite and aircraft microwave
measurements for the development of large-scale soil moisture
retrieval algorithms. In addition, considering SMOS Cal/Val
activities and the pre-launch assessment of SMAP, CanEx-
SM10 is one of the few soil moisture experiments conducted
over a boreal forest and subarctic environment.
The following sections describe the CanEx-SM10 study
sites and the experimental design including the sampling
strategy associated with the ground and airborne
measurements and the selection of SMOS and other satellite
acquisitions. The analysis of data collected during CanEx-

SM10 is then presented followed by a short description on the Fig.1. The CanEx-SM10 study area including both the Kenaston agricultural
CanEx-SM10 data base. site and BERMS boreal forest site.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING 3
2.1.1. Kenaston site 2.2. Ground truth locations
The agricultural site (51.50 N, 106.50 W) is located Measurements to characterize the soils and vegetation were
approximately 80 km from Saskatoon (Fig. 1), Saskatchewan, spatially distributed over the Kenaston and BERMS sites,
Canada. The topography of the region is shown in Fig. 2. As capturing the natural variability in the landscape. Sampling
evident in this figure, the region is not perfectly flat and the stations were selected based upon availability of resources,
highest elevations are in the eastern part of the area and there road accessibility, and ability to meet two objectives: 1)
is a valley toward the west. Based on Landsat image provide a suitable dataset to validate satellite and airborne soil
classification, approximately 92% of the site is under annual moisture retrieval; and 2) relate point measurements to
cultivation with most of the remaining area in permanent grass satellite acquisitions. A total of 60 fields were sampled over
and pasture. Production is almost exclusively rain-fed with the Kenaston site. These included 24 fields instrumented with
minimal use of irrigation. Prior to and during the CanEx- long term in situ soil moisture stations managed by EC and 16
SM10 experiment, the Kenaston area experienced above fields instrumented managed by U of G. An additional 20
normal rainfall resulting in very wet soil conditions. As a fields were added to complement these permanents sites
consequence, pools of standing water were present in many (Table I and Fig. 2).
fields, increasing the percentage of open water from 1.5% to Over the BERMS site, the 35 sampling stations consisted of
4.9 % [13], [14]. The presence of standing water inevitably 6 BERMS permanent stations located at BERMS research
complicates the analysis and interpretation of the coarse sites, 20 BERMS temporary stations installed by the U.S.
resolution microwave signals. Department of Agriculture (USDA) from May to August
Table I describes the field conditions during CanEx-SM10. 2010, and several ground truth sites. All of these 35 stations
With the exception of the pasture fields, most fields had been were sampled on the airborne flight day [13].
tilled and were covered with varying amounts of crop residue.
Seeding and crop development were delayed in the spring of
2010 due to the unusually wet soil conditions. Vegetation III. EXPERIMENTAL DESIGN
cover varied but was less than 50% for most fields (Table I).

Although most of the soils are loamy, high variability was
CanEx-SM10 was an intensive short-term campaign (31
observed in soil texture, and the dominant textures included
May to 16 June, 2010) designed to collect consistent field
silt, clay and sandy loams.
measurements at a time close to satellite and airborne
acquisitions to support validation of both SMOS and SMAP

products. Table III presents a comprehensive list of the field
data collected during CanEX-SM10. The spatial extent of both
the Kenaston and BERMS sites was equivalent to about two
SMOS pixels. The size of the study sites impacted the
experimental design and was a factor in optimizing the
number of sampled stations. This optimization included
minimizing sampling times and travel time from one field to
another as well as coordinating sampling to be coincident with
SMOS overpasses, all within available resources. Given these
constraints and the requirement to collect spatially distributed
soil and vegetation measurements (moisture, roughness,
Fig.2. Digital Elevation Model of the Kenaston site at 30 m resolution
(downloaded from http://www.geobase.ca/) along with the location of the biomass, LAI, bulk density, etc.), the priority was to cover a
sampling stations. Basic information for all the stations is provided in Table large number of fields at the SMOS scale of approximately 30
1. km. A calendar of data collection and information on the

available airborne and satellite acquisitions are provided in
2.1.2. BERMS site Table IV.

The BERMS region is located north of Prince Albert (53.24 3.1. Ground data sampling strategy
N, 105.75 W) in Saskatchewan near the southern extent of the
boreal forest (Fig. 1). BERMS features several instrumented 3.1. 1. Soil moisture, temperature and bulk density
research sites located in regions with various vegetation types
Over the Kenaston area soil moisture, bulk density, and
(mostly forest), ages and structures [15]. The topography is
temperature were measured approximately coincident with the
generally rolling and the dominant vegetation type depends on
satellite and airborne acquisitions. On each sampling day,
the soil types and drainage conditions. To reduce the
measurements were taken on 48-60 fields, with each team of
contribution of lakes to the observed microwave signals,
two visiting four to five fields. The location of each sampling
CanEx-SM10 only covered the eastern portion of the study
point in each field was recorded using a GPS. During
area which was originally defined for the BERMS
subsequent sampling days, these coordinates were used to
measurements program [13]. Five vegetation types (old
navigate to the same point, ensuring that each successive
Aspen, old Jack Pine, Harvested Jack Pine, Fen, and old Black
measurement was taken at the same location.
Spruce) mostly forests are dominant in this region (Table II).
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING 4
TABLE I
CHARACTERISTICS OF THE SAMPLING FIELDS OVER KENASTON AREA - 24 EC FIELDS, 16 U OF G FIELDS, AND 20 MANUAL SURVEY (MS)
FIELDS

Crop Soil
Field Field Lat Long
Fract. R esidue Bulk
type # (°N) (°W) Type Height Sand Silt Clay Type Tillage
cover cover density
- cm % % g/cm³ % % %
EC D4 51.25 106.45 Canola - >5 0 1.18 26 51 23 Silt Loam Yes
H1 51.39 106.52 Canola - 5-11 <25 1.18 42 41 17 Loam No

H2 51.37 106.50 Canola - 8 >50 1.12 39 44 17 Loam No

H3 51.36 106.51 Canola - 12-35 >50 1.17 34 50 16 Loam No
H4 51.37 106.45 Cereal 2 25 -50 1.14 50 32 18 Loam Yes
H5 51.38 106.43 Canola 1 15 25 -50 1.19 37 41 22 Loam No
I1 51.37 106.43 Wheat - - <10 0.87 28 59 13 Silt Loam No

I2 51.38 106.42 Wheat 8-10 1 <25 1.20 45 22 33 Clay Loam No

I3 51.39 106.41 - - - <10 0.98 28 53 19 Silt Loam No
I4 51.38 106.41 - - - <1 0 0.92 29 53 18 Silt Loam No
J1 51.39 106.45 Wheat - 9 >5 0 1.19 23 57 20 Silt Loam No
J2 51.40 106.43 Wheat - 18 >5 0 1.31 26 50 24 Loam Yes
J3 51.39 106.43 Canola - 7-15 >50 1.34 31 46 23 Loam No

J4 51.41 106.43 Canola 4 11 >50 1.32 37 40 23 Loam No
J5 51.42 106.42 Wheat - 3-7 >5 0 1.30 33 46 21 Loam No
K1 51.42 106.42 Wheat 3 - >5 0 1.14 29 49 22 Loam Yes
K2 51.43 106.43 Peas - 2 >5 0 1.09 20 43 37 Clay Loam No
K3 51.44 106.43 Canola 1 9-17 >50 1.06 33 47 20 Loam No

K4 51.44 106.43 Peas 3 7 >50 1 21 53 26 Silt Loam No
K5 51.45 106.50 Canola 3 6 >5 0 1.16 - - - - Yes
L1 51.43 106.47 Wheat 25-30 60-80 <2 5 1.11 26 55 19 Silt Loam No
L2 51.42 106.47 Wheat 10-15 40 <2 5 1.13 29 49 22 Loam Yes
L3 51.43 106.54 Wheat 6-8 10-13 <25 1.09 34 45 21 Loam No

L4 51.45 106.57 Canola 4-5 15 >50 1.03 37 42 21 Loam Yes

U of G A2 51.57 106.18 Wheat 8-10 9-38 >50 1.26 29 42 29 Clay Loam No
A3 51.63 106.10 Pasture 10-25 38 25 -50 1.13 - - - - No
A4 51.59 106.01 No crop - - - - - - - -
A5 51.54 105.99 Peas 2 5-10 25-50 1.10 45 31 24 Loam No

C1 51.36 105.94 Pasture 20-25 50-73 >10 1.25 41 38 21 Loam No

C2 51.39 106.10 Not planted - - >50 1.08 41 38 21 Loam No
C3 51.43 106.24 Wheat 8-10 8-14 <2 5 1.24 32 40 28 Clay Loam No
C5 51.37 106.29 Wheat - 8-10 >5 0 1.03 20 53 27 Clay Loam Yes
E1 51.27 106.39 Pasture - - <2 5 1.24 25 54 21 Silt Loam No
E4 51.36 106.41 Bare soil - - >50 1.23 23 59 18 Silt Loam No

F2 51.28 106.67 Canola - 9-15 25-50 1.10 30 49 21 Loam Yes
G2 51.36 106.63 Lentil 2-3 11 >5 0 1.21 34 48 18 Loam Yes
G4 51.36 106.57 Peas 3-4 15 >5 0 1.14 31 52 17 Silt Loam No
G5 51.39 106.50 wheat - 13 >5 0 1.15 28 47 25 Loam No
L5 51.56 106.24 Lentil - 7 >50 1.31 38 43 19 Loam No

I5 51.40 106.45 - - - <10 1.02 30 54 16 Silt Loam No

MS A1 51.56 106.24 No crop - - >50 1.12 40 39 21 Loam Yes
A6 51.42 105.94 No crop - - 25 -50 1.12 35 38 27 Clay Loam No
B1 51.42 105.94 Pasture - 39 <1 0 1.33 58 28 14 Sandy Loam No
B2 51.42 105.90 Pasture - 54 <10 1.34 72 17 11 Sandy Loam No

B3 51.42 105.87 Pasture 10-40 23-75 <10 1.12 50 39 11 Loam No

B4 51.41 105.86 Pasture - 28-50 <10 1.21 56 30 14 Sandy Loam No
B5 51.50 106.09 Pasture - 7-22 >5 0 1.04 - - - - No
C4 51.40 106.24 Bare soil - - >5 0 1.23 33 41 26 Loam Yes
D1 51.39 106.29 Lentil 2-3 4 >50 1.16 30 43 27 Clay Loam Yes

D2 51.44 106.33 Canola - 6 25-50 1.10 37 42 21 Loam Yes

D3 51.42 106.45 wheat 4 11 25-50 1.18 34 39 27 Clay Loam Yes
E2 51.33 106.36 Bare soil - - >5 0 1.32 31 51 18 Silt Loam No
E3 51.33 106.39 Lentil - - >5 0 1.19 23 54 23 Silt Loam No
E5 51.28 106.61 Wheat 4-8 7 25 -50 1.26 28 51 21 Silt Loam Yes
F1 51.27 106.66 lentil 3 9 >50 1.21 42 40 18 Loam No

F3 51.29 106.68 Not planted - 8 >50 1.27 37 45 18 Loam No
F4 51.30 106.66 Not planted - - >5 0 1.23 39 47 14 Loam No
F5 51.39 106.78 Lentil 3 20 25 -50 1.17 38 46 16 Loam Yes
G1 51.38 106.61 Not planted - - >5 0 1.25 33 49 18 Loam No
G3 51.33 106.67 Not planted - - >50 1.13 32 46 22 Loam No
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING 5
In each field, soil moisture was measured to a depth of 6 cm
TABLE II
using the Steven’s Water Hydra Probe inserted vertically.
DESCRIPTION OF THE FIVE BERMS SAMPLING SITES
Sampling was conducted along two transects 400 m apart.
Each transect included seven sample points at a 100 m
Geographic
Site_ID location Site description spacing. At each sample point, three replicate moisture
Mature wet old black spruce readings were collected. When tillage structure was evident,
53.99 N
OBS Moss and Labrador tea understory these replicates were located at the top, bottom and side of the
105.12 W
Mean height: 7 m tillage furrow. Table V presents the sampling regime for soil
Old dry jack Pine
moisture, soil temperature, thermal infra red (TIR), and bulk
mature dry coniferous
53.92 N density at Kenaston. For each field and on each sampling day,
OJP jack pine forest
104.69 W a gravimetric sample was obtained for a fixed volume of the
Lichen understory
Mean height : 13.4 m surface layer. These samples were taken to the laboratory for
Harvested Jack Pine 2002 oven drying over a 24 hour period. Then, they were used to
HO2 53.95 N Ground cover consisting of sparse grass, calibrate the soil moisture probes and to derive soil texture
104.65 W shrubs and immature jack pine seedlings
and bulk density via lab analysis.
Mean height: 1.82 m
Mixed Forest In addition to the manual sampling of soil moisture within
53.90 N Pine, Fir, Aspen each field, hourly soil moisture and soil temperature profiles
Temp 7
104.88 W Mean height: 6.44 m, 6.87 m, and 10.17 m at 5, 25, and 50 cm depths were recorded continuously at
for respectively Pine, Fir and Aspen
single points by the EC and U of G networks. They also used
Flooded vegetation, among others horse tail,
the Steven’s Water Hydra probes installed vertically and
grass and 2-3 kind of shrubs.
53.78 N
FEN Mean height: 35.6 cm, 45.7 cm, and 43.2- horizontally for respectively EC and U of G networks. Using
104.62 W
96.5 cm for respectively horse tail, grass and the calibration curves developed for each network station,

shrubs.
uncertainty in volumetric soil moisture ranged from +/- 0.03
m3/m3 to +/- 0.015-0.02 m3/m3, depending on the soil texture
[16]. Some additional details regarding the network operated
TABLE III
by the U of G are described in [17]. These profiles of soil
GROUND DATA COLLECTED AT THE KENASTON AND BERMS
moisture and soil temperature were complemented by
SITES
precipitation measurements from rain gauges.
Human Sampled Over BERMS, a one-day field campaign was conducted on
Sites Measurements
resource stations 16 June 2010. Soil moisture, bulk density, and temperature
soil moisture at 6-cm
measurements were collected approximately coincident with
Kenaston depth, bulk density,
48-60 per the aircraft and SMOS acquisitions. In the sampling approach
soil temperature at 5 12 teams of
sampling
and 10-cm depth, and 2 people for this site, measurements were taken at 35 ground truth
day
Thermal Infra Red stations (GTS) that were spatially distributed over the study
(TIR)
area and located along accessible roads and trails [13]. At
June 1-14, 2010 60 for the
12 teams of each GTS station, three soil moisture measurements were
Soil texture entire
2 people campaign taken at a 6-cm depth and at three measurement points located
Vegetation (water within the surrounding canopy at a nominal distance of 20, 25,
60 for the
content, height, 2 teams of entire and 30 m from the GTS location. The sampling was
density, etc.) and soil 2 people
campaign conducted by six teams of two people and covered the entire
roughness
area, within the limits of road inaccessibility. Table V
60 for the
Leaf Area Index (LAI) 3 people entire presents the sampling strategy for soil moisture, soil
campaign temperature, TIR, and bulk density at BERMS. As at the
BERMS Soil moisture, soil 35-40 for Kenaston site, pre-programmed GPS coordinates were used to
temperature at 5 and 6 teams of the
easily and accurately geo-locate the sampling stations. In
16 June, 2010 10-cm depth, surface 2 people sampling
temperature and TIR day some cases, the collection of bulk density samples over
Tree characteristics BERMS was complicated by the presence of an organic layer
(DBH, height, crown of variable thickness. At each of the three replicate sampling
fractional cover, stem
locations, the organic layer was first measured and then
14-16 June and density, branch 1 team of 5
on 13-20 July, measurements, soil people 5 removed from a 20 cm x 28 cm area in order to collect a
2010 moisture at 5 cm sample of the underlying mineral soil from which the bulk
depth, stem and branch density was derived. The depth of the organic layer was
dielectric constant)
recorded and the material bagged and weighed for the
Understory
1 team of 5 determination of water volume.
characteristics (type, 5
people
fractional cover)
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING 6
TABLE IV
AVAILABLE GROUND, AIRBORNE, AND SATELLITE MEASUREMENTS DURING CANEX-SM10

Measurements Sites Kenaston BERMS
June 2010 1 2 3 4 5 6 7 8 9 10 11 12 13 14 16
Ground Data Collection - √ - - √ √ √* - √ - - - √ √ √
Satellite SMOS √√ √ √√ - √√ - √ √√ - √√ √ √ √√ - √√
AMSR-E √√ √√ √√ √ √√ √√ √√ √√ √√ √√ √√ √√ √√ √√ √√
RADARSAT-2 √√ √ - - √√ - - √√ - - √ √ - - -
ASAR - - - - - √ √ - - √√ - - √ - √
ALOS-PALSAR - - - - - √ √ - √ - - - - √ -
Airborne Twin Otter and
- √ - - √ √ √* - √ - - - √ √ √
UAVSAR

√: one acquisition per day
√√: two acquisitions (ascending and descending) per day
√*: Partial coverage due to rain event

TABLE V
SAMPLING REGIMES OVER THE KENASTON AND BERMS STUDY SITES

Sites Soil sampling regime
Sampling points per field: 14 located at pre-programmed GPS points
Transects per field: 2 transects 400 m apart
Points per transect: 7
Spacing between points in transect: 100 m with the first point 50 m from the field edge
Number of soil moisture readings per point 3 (top, bottom and side of furrow)
Kenaston Soil moisture measurements Probe inserted vertically, soil moisture is integrated over 6 cm
Soil temperature 4 points at two depths (5 cm and 10 cm)
Thermal infra red (TIR) 4 measurements in each field. Exposed Vegetation, shaded vegetation, exposed ground
and shaded ground.
Bulk Density 1 core sample of 5 cm depth
Site Photos One taken in the direction of the crop row,
Sampling points per GTS: 3, located at pre-programmed GPS points
Spacing between points: 5 m (20 m, 25 m and 30 m from the GTS)
Number of soil moisture readings per point 3 (top, left and right side of measurement point)
Soil moisture measurements Probe inserted vertically, soil moisture is integrated over 6 cm
BERMS Soil temperature Simultaneously to soil moisture at 5 cm depth
Thermal infra red (TIR) 4 Measurements for each GTS. exposed vegetation, shaded vegetation, exposed Ground
and shaded ground.
Gravimetric soil moisture 3 samples of 5 cm depth per GTS
Site Photos Two landscape and one vertical
Vegetation sampling regime
Vegetation characterization plant density, row spacing, row direction
Wet and dry biomass and canopy water content 1 m sampling if rows were well defined, otherwise
sample of 50 cm x 50 cm using a gridded board, 3 replicates. Wet samples oven dried t
Kenaston determine dry biomass and canopy water content.
LAI 14 hemispherical photos along 2 parallel transects 30 m in length
Site photos 1 photograph of a gridded board placed over the vegetation, 3 replicates;
14 crop architecture photos
Vegetation height and stem diameter 3 – 10 height and diameter measurements per site at each of three sites
Transects per site: 3 of 100 m length for the mixed forest (Temp 7) and 1 of 100 m length for OBS, OJP, an
HO2 sites
Spacing between points in transect: 10 m
Vegetation characterization Species identification, tree height, diameter-at-breast-height (DBH), and tree count, crow
BERMS
depth
Densities of Stems, large and small branches, leaves From tree trunk density
Radius and length of large and small branches, leaves From a destructive sampling of one ‘average’ tree
Distribution parameters From photographs and site inspection
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING 7
The above data sets were augmented with soil moisture and at 80 oC to constant weight, which provided both dry biomass
soil temperature which are continuously collected at different weights and canopy water content. Due to time constraints,
depths and at 4-hour intervals at the permanent BERMS each field was sampled once for the derivation of the
research stations [15]. The only exception was the Fen site abovementioned parameters.
where data were recorded every 30 minutes. Furthermore, In addition to destructive vegetation sampling, crop
over BERMS, CanEx-SM10 also benefited from 5-cm depth development was also monitored with the measurements of
soil moisture measurements collected on an hourly time LAI. At each site, a total of 14 hemispherical photos were
interval at 20 temporary stations [13]. taken at 5 meters spacing along two parallel transects
approximately 30 meters apart. This method of LAI
3. 1.2. Soil roughness determination was well suited conditions in this experiment
given the limited canopy development. Coincident with the
The soil roughness measurements were made over the
LAI measurements and accompanying each set of
Kenaston fields using a 1-meter pin profiler consisting of 200
hemispherical photos, crop architecture photos were also
needles spaced at an interval of 5 mm. Each field was sampled
collected at each site. A summary of the sampling of
at least once, however resampling was conducted over fields
vegetation characteristics is given in Table V.
that were tilled during the campaign. The objective was to
BERMS data will be used to investigate how well soil
measure soil roughness characteristics (standard deviation of
moisture can be retrieved in boreal landscapes using L-band
surface heights and correlation lengths) to quantify the impact
active/passive microwave remote sensing. The data will also
of roughness on SAR backscatter and to a lesser extent on L-
assist in improving SMOS soil moisture retrieval algorithms,
band passive microwave data for estimating soil moisture at
in developing SMAP soil moisture retrieval algorithms and in
the SMOS scale. Due to the expanse of the study area (about
forward modeling of SMAP radar backscatter. At BERMS a
two SMOS pixels), an approach was adopted to optimize the
total of five sites were sampled (Table II). The ground
number of roughness measurements across the site. Data
measurements included three 100-m transects at a mixed
collected in July of 2008 over Kenaston was analyzed to
forest site (Temp7) and one 100 m transect at each of the Old
determine the within field variance in surface roughness to
Jack Pine, Old Black Spruce, and Harvested Jack Pine sites.
guide the sampling design. This analysis determined that the
The Fen site vegetation characteristics were measured along
within field variance in roughness is far less than the field to
the boardwalk leading to the flux tower location. Various
field variance. Roughness in agricultural regions is largely
vegetation measurements were taken in 10-m intervals along
driven by tillage applications and thus this observation is not
each transect. At every 10 m mark, tree height, trunk radius,
unexpected. Based on this analysis it was determined that one
and tree count were measured together with trees fractional
sample site per field was sufficient to characterize roughness.
cover, understory cover, and litter depth. The stem density
The pin profiler is positioned perpendicular to the soil and
along the entire transect was determined by counting the
once the board is level, the needles are released. The tops of
number of stems within a ~2 m arm-span and dividing by the
the needles mimic the surface roughness profile. At each site,
area (approximately 100 m x 2 m). The densities of large and
a 3 m roughness profile was created by placing the one metre
small branches as well as that of leaves were calculated from
profiler end to end in the look directions of both the UAVSAR
the trunk density and the quantity of these components for the
and RADARSAT-2 (descending overpass). This 3 meter
measured trees at each 10 m mark. Crown layer depth and
profile was replicated three times, at a distance of
trunk height, as well as trunk diameter-at-breast-height (DBH)
approximately five metres. A digital camera recorded the pin
were recorded. For each forested site, one “average” tree was
meter profiles and these photos were processed to derive
destructively sampled from which the radius and length of
surface roughness characteristics (standard deviation of
large and small branches as well as leaf dimensions were
surface heights and correlation lengths). Processing of the
recorded. The distribution parameters of the branches were
photos and the extraction of the roughness statistics are
deduced from photographs and inspection in the field.
described in [18]. The mean and the standard deviation of the

surface roughness parameters were computed to determine the
3.2. Remote sensing data
average field roughness.
Over the BERMS forested site no roughness measurement To meet the objectives of CanEx-SM10, both airborne and
was collected due to the presence of an understory. satellite remote sensing data were acquired.

3.1.3. Vegetation 3.2.1. Aircraft data
The Kenaston data will be used to assess the impact of Two aircraft, one equipped with a passive microwave
canopy water content on the microwave response in radiometer and the other with an active SAR, were used in
estimating soil moisture at SMOS/SMAP scales. For each CanEx-SM10. These included a Twin Otter aircraft owned by
field, three replicate vegetation samples were gathered at a the NRC and managed by EC, and NASA’s Gulfstream-III
single site. Measurements of plant height, stem diameter, plant (G-III) aircraft. These aircrafts were deployed to acquire data
density, row spacing and row direction were recorded. To to support the validation of SMOS products (L1, L2), the pre-
minimize crop disturbance, vegetation in front of the 1-meter launch assessment of SMAP data, and evaluation of soil
pin profiler was removed, providing a measurement of above moisture retrieval algorithms from these two missions. The
ground wet biomass. The vegetation samples were oven-dried data will also be used to investigate approaches to scaling
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING 8
among remote sensing sources and to understand the IV. PRELIMINARY RESULTS
relationship between ground measurements and satellite
products. The Twin Otter and G-III attempted to cover the 4.1. Ground measurements
Kenaston and BERMS study areas close in time to SMOS
4.1.1. Soil moisture, bulk density and temperature
overpasses. The flight calendar is presented in Table IV.
Over the Kenaston fields, a site specific calibration of the
Twin Otter: This aircraft was equipped with EC’s passive volumetric soil moisture measured by the Steven’s Water
microwave radiometers which operate at 1.4, 6.9, and 19-37- Hydra Probes was performed using the gravimetric soil
89 GHz. Visible and infra red radiometers were also mounted samples. Fig. 3 shows a summary of these dataset and
on the aircraft and these sensors provide variable spectral suggests a strong agreement between soil moisture measured
information suitable to assist with data analysis and modeling. by the two methods. The data spread observed around the 1:1
About 16 parallel flight lines were required to cover each line can be attributed to variances in soil type and errors in
study area. The L-band radiometer was flown at an altitude of collecting gravimetric samples and thus in estimating soil bulk
approximately 2.3 km which resulted in a spatial resolution of density. The soil bulk density values derived from the
about 2.25 km. These L-band data were collected at a 40° gravimetric samples are presented in Table I for each field.
incidence angle.
0.6

NASA G-III: This aircraft carried the UAVSAR which is a bias=0
r
fully polarimetric L-band radar [19]. Using multiple flight te 0.5 RMSE=0.05
lines, the UAVSAR provided spatial coverages similar to a
w R²=0.67
those of the L-band radiometer with a nominal flight altitude c
of 13 km. The UAVSAR collected data over a swath of about etri³) 0.4
21 km with the incidence angle ranging from 20° (near range) mm
u³/
to 65° (far range). The pixel size is 7.5 m in range x 6 m in olm 0.3
azimuth. The UAVSAR data are publicly available from the d vnt (
UAVSAR data server of the Jet Propulsion Laboratory [19] ee
for both the Kenaston and BERMS sites. suront 0.2
ac
e
Full details on the flight lines of both the Twin Otter and m
the G-III as well as additional information on passive and d 0.1
el
active microwave sensors aboard these aircrafts can be found Fi
in the Experimental plan of CanEx-SM10 [13]. 0
0 0.1 0.2 0.3 0.4 0.5 0.6
3.2.2. Satellite data Hydra Probe measured volumetric
SMOS acquisitions available over the study sites during water content (m³/m³)
CanEx-SM10 are listed in Table IV. Other satellite
Fig.3. Calibration curve of the hydra probe sensors over the Kenaston fields.
acquisitions (AMSR-E, RADARSAT-2, Envisat ASAR and

ALOS-PalSAR) were planned to be as close in time as
The individual field average soil moisture measured at
possible to the SMOS overpasses. Several modes of
Kenaston during CanEx-SM10 are presented in Fig. 4a. The
RADARSAT-2 were planned including acquisitions of Fine
high soil moisture values reflect the very wet conditions due
Quad Polarimetric, Standard and Wide Swath, at varying
to heavy rainfall before and during the field campaign
incidence angles. Envisat ASAR acquisitions in Alternating
(Section 2). Some variation in soil moisture is observed
Polarization and Wide modes were programmed to fill gaps in
between fields. A number of factors contribute to inter-field
the RADARSAT-2 acquisition plan. ALOS-PalSAR data were
differences in wetness including topography, precipitation
acquired in Fine Dual and Wide modes. To maximize
amounts, soil texture, and vegetation cover (Fig. 2 and Table
temporal coverage, whenever possible both ascending and
I) and will be explored in greater detail in the future. As a
descending microwave acquisitions were programmed. L- and
complement of Fig. 4a, the temporal evolution of the averages
C-band microwave satellite data (Table IV) will be compared
soil moisture, soil temperature and precipitation data is given
with L- and C-bands airborne data to understand the scaling
in Fig. 5. The lowest soil moisture conditions were observed
effect on soil moisture and to develop active/passive soil
for June 2nd, 5th and 6th. Rain on June 7th and 8th resulted in
moisture retrieval algorithms.
very wet conditions on June 9th. Warm and dry conditions
In addition to microwave satellite data, LANDSAT, SPOT,
observed after June 12th led to the soil drying towards the
and AWiFS optical measurements were available over the
conclusion of the experiment. Indeed, soil moisture values on
Kenaston site.
June 13th and June 14th were lower than the values observed

on June 9th. Fig. 6 shows the coefficient of variation as a
function of the mean soil moisture values measured during
CanEx-SM10. These statistics indicate a decrease in the
relative variation in soil moisture with an increase in moisture.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING 9
-a-
0.5
m³) 0.4
m³/
e (
ur 0.3
st
oi
m
oil 0.2
S
June 2 June 5 June 6 June 9 June 13 June 14
0.1
A1A2A3A4A5A6B1B2B3B4B5C1C2C3C4C5D1D2D3D4E1E2E3E4E5F1F2F3F4F5G1G2G3G4G5H1H2H3H4H5I1 I2 I3 I4 I5J1J2J3J4J5K1K2K3K4K5L1L2L3L4L5
Fields
-b-
35
C)
m (° 30 June 2 June 5 June 6 June 9 June 13 June 14
5 c 25
e at 20
ur
at 15
er
p
m 10
e
oil t 5
S
0 A1A2A3A4A5A6B1B2B3B4B5C1C2C3C4C5D1D2D3D4E1E2E3E4E5F1F2F3F4F5G1G2G3G4G5H1H2H3H4H5I1 I2 I3 I4 I5 J1J2J3J4J5K1K2K3K4K5L1L2L3L4L5
Fields
-c-
4
m)
ht (c Profiler in the look direction of RADARSAT-2 Profiler in the look direction of UAVSAR
g 3
ei
h
s
s
ne 2
h
g
u
o
e r 1
c
a
urf
S
0
A1A2A3A4A5A6B1B2B3B4B5C1C2C3C4C5D1D2D3D4E1E2E3E4E5F1F2F3F4F5G1G2G3G4G5H1H2H3H4H5I1 I2 I3 I4 I5 J1J2J3J4J5K1K2K3K4K5L1L2L3L4L5
Fields
Fig.4. Field average soil characteristics over Kenaston during CanEx-SM10 a) 6-cm soil moisture; b) 5-cm depth soil temperature, and c) surface root mean square
(rms) roughness height

40
June 2
June 5
0.6 60 )
moisture (m³/m³) 00..24 MMReaeaainssfauulrrleedd ssooiill mteomispteurraet ure 2400 n volume (mm)emperature (°C) nt of variation (% 2300 JJJJuuuunnnneeee 691134
oil Raioil t cie
S S effi 10
o
0 0 C
115500 115555 116600 116655 117700
Julian day 2010

Fig.5. Temporal evolution of the mean values of the measured soil moisture, 0 0.1 0.2 0.3 0.4 0.5
soil temperature and precipitation over Kenaston during CanEx-SM10
Measured soil moisture (m³/m³)

Fig.6. Coefficient of variation in soil moisture versus the mean values of the
measured soil moisture (m³/m³) during CanEx-SM10
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING 10
Famiglietti et al. [20] have shown that this decrease in Measurements and vegetation observations collected over
variance is reduced at higher moisture levels, within a range BERMS are described in Table III. Tree heights varied from
of 0.20-0.45 m3/m3 soil moisture. This suggests that other site 1 to 22 m; tree heights were greatest at the OJP site (8-19 m)
factors may play a role and thus might explain the scatter followed by the OBS site (2-13 m). Younger trees dominated
observed in Fig. 6. The field to field variation of some of these the HO2 with tree heights varying from 1 to 3 m. In Table VI,
factors is given in Table I. Current studies are focused on the strong relationship between tree heights (in meter) and the
attributing the observed variance to physical processes. diameter-at-breast-height (DBH in meter) is demonstrated for
Soil temperature was measured at a 5-cm depth sites Temp 7, HO2, and OBS. A much weaker relationship is
simultaneous with the soil moisture measurements. The observed for the OJP site (Table VI).
temporal trend in soil temperature matches that of soil
moisture as presented in Fig. 4a. The soil temperature ranged TABLE VI
RELATIONSHIPS DBH (D IN M) VS TREE HEIGHT (H IN M)
from 10 to 20°C.
MEASUREMENTS AT DIFFERENT SITES OF BERMS

4.1.2. Soil roughness Sites Tree species Linear relationships
Temp 7 Pine (only) D=0.012H−0.009 ; R2 = 0.81
The measurements of soil roughness in the look direction of
OBS Old Black Spruce D=0.011H−0.002 ; R2 = 0.80
both RADARSAT-2 (91° in descending) and UAVSAR
HO2 Harvested Jack D=0.02H−0.012 ; R2 = 0.84
(242°) are shown in Fig. 4c. In some cases, there was no Pine
significant macro tillage structure and the two measurements OJP Old Jack Pine D=0.0074H+0.038 ; R2 = 0.41
were similar. In the fields with tillage structure, roughness did OJP+HO2 Old and harvested D=0.0096H+0.0069 ; R2 = 0.94
vary as a function of the SAR look direction. In addition, Jack Pine
roughness measured in the look direction of the UAVSAR All All D=0.0096H+0.0065 ; R2 = 0.82
was higher than that measured in the look direction of
RADARSAT-2. Consequently, a constant surface roughness
cannot be assumed in backscatter modelling in this region.
4.2. Remote sensing data
4.1.3. Vegetation
4.2.1. Aircraft
Several vegetation characteristics were measured over the
4.2.1. UAVSAR
Kenaston and BERMS sites (Section 3.1.3). In this paper, the
consistency of the data was evaluated empirically. For the The UAVSAR acquired data over the Kenaston site at
Kenaston fields, Fig. 7 demonstrates a positive relationship incidence angles of 20-65 degrees. The original images were
between LAI and percent crop fractional cover. This is processed to produce a normalized data set with an incidence
expected during early crop development, which was the angle of 40° [21]. Fig. 8a is an R-G-B (HH-HV-VV) color
CanEx-SM10 condition. At higher LAI, this relationship composite of 13 June, 2010 acquisition. The extent of the
weakens as crop cover becomes near complete yet LAI UAVSAR coverage and its location within the Kenaston site
continues to increase. Additional information on crop are provided (upper left corner of the image) in Fig. 8a. In this
characteristics associated with the Kenaston fields are given in figure, the strongest response is observed for the HH (red
Table I. color) and VV (blue color) polarizations, with much lower
contributions from the HV (green color) channel. This
indicates a dominance of surface scattering from bare or
3.5 sparsely vegetated surfaces, with little contribution from
Canola volume scattering. Variations in the HH and VV responses are
3 Lentile
evident and these reflect the field to field differences in soil
Pasture
Wheat moisture and roughness (Fig. 4a and Fig. 4c). Very dark
2.5
Peas locations are often associated with specular reflection from
2 standing water. The UAVSAR data adequately discriminates
AI between the different terrain features present at the Kenaston
L
1.5
site and captures the ground conditions (i.e. soil moisture,
vegetation cover) during the campaign.
1
The UAVSAR data acquired over the BERMS site on 16
0.5 June 2010 can be seen in Fig. 8b where the individual data
swaths with 25- 65° incidence angle range were post-
0
processed by geo-referencing and assembling them into a
0 20 40 60 80 100 single image mosaic to cover the whole area of interest. The
Crop fractional cover (%) image is an R-G-B (HH-HV-VV) color composite of 16 June

2010 acquisition. The location of the five BERMS sites

Fig.7. Measured effective LAI versus percent crop fractional cover sampled for vegetation can be identified by their location with
respect to White Gull Lake, which shows up prominently in